Fiber laser welding penetration depth monitoring: A novel method using plume visual and SMI signal fusion analysis

The penetration depth is an important indicator for evaluating the laser penetration ability in the laser welding process. Plume is the main information carrier of welding process signals. In this paper, by synchronously collecting the plume vision signal and the plume particle signal (based on the principle of self-mixing interference, SMI), and combining the algorithms of ensemble empirical mode decomposition and fast Fourier transform (EEMD-FFT) to extract the time-domain and frequency-domain features of the plume signals, a method designed for penetration depth monitoring through the fusion of multiple plume signals is introduced. The results show that both the plume area and the total intensity of the SMI signal are positively correlated with the weld penetration depth, and the frequency-domain features of the plume signal have a higher correlation than the time-domain features. Compared with the peak frequency, the centroid frequency of plume signal has higher sensitivity and adaptability to working conditions when reflecting the changes in the penetration depth. The SMI signal has obvious advantages in signal processing efficiency. Its storage space is only 0.47 % of that of visual signal, and the overall processing time can be shortened by 93.4 %. The two types of signals have good complementarity in terms of information. The signal strategy can be flexibly selected according to actual needs to achieve a balance between the efficiency and accuracy of welding monitoring. The research can provide a novel technical scheme for the in-situ monitoring during laser welding.